Evolving law, Smolin and others

[friend]

Again, a fundamental principle in physics is that whatever's foundational needs to be "in only one form" and hopefully fairly simple. I think Fra would agree, but I'm not sure I do, because I'm not sure any information can be meaningfully defined unless there are different kinds of information out there in the environment.

Conrad

Seriously, information can only be calculated about some other structure. It never exists in a vacuum. But can co-exist with that stucture. It is not fundamental in that the underlying structure must exist first before any information can be calculated about it.

I'm thinking in terms of purely mathematics. How does one calculate information from nothing? But then again, maybe it IS possible to "derive" structure from its information content. That sounds like a harder thing to do.

For example, it is very easy to show how the Path Integral from QM can be derived solely from the Dirac Delta function. See:

http://hook.sirus.com/users/mjake/delta_physics.htm

But the Dirac Delta function is itself a distribution about which information can be calculated. Now, if it should turn out that all of physics can be derived from a Dirac Delta function, and that the information of the Delta function is constant, for example, then maybe we can derive a law of conservation of information and then laws of physics from that. Who knows?

 

[Fra]

Well, I've already made clear I think you're on the right track, for whatever that's worth.

Yes you do seem to connect to the main reasoning and I appreciate your feedback.

But "selected for stability" doesn't sound quite right, maybe because to me that seems to imply a background time structure (even if not a time-metric). In biology, the evolutionary game is all about how to get complex systems to last through time (by copying them before they inevitably break down due to their complexity). But in physics, the problem seems to be different, namely how to define any information in a meaningful way in the first place -- how to get any information to make a difference to anything. Time maybe comes about only in this process.

This is hard to write about but I tried to describe that different observers might see different degrees of freedom, and that this conflict when the observers interact evolves their memory structures. Here I associate physical interaction with negotiation. The negotiation imposes a mutual selective pressure to find an agreement. In that sense consensus is emergent.

The Shannon theory is of course important to physics because it's merely quantitative -- it abstracts from all questions about what information "means" (how it affects things, how it's measured), and that's good if physics doesn't yet have ways of dealing with those questions. But I agree with you entirely that we need an information theory that explains why and how information does what it does -- i.e. gets defined / determined and in turn provides a context that defines / determines other information.

One objection to shannons entropy is that it is often treated like a universal measure of information, but it is in fact relative to the choice of a equiprobable microstructure. I argue that this CHOICE must not be treated as a theorists armchair maneuver, I think there is physics behind the choice, and this choice is evolving, because the microstructure is a dynamical entity formed by interactions.

On the other hand, the connection between stability of information and mass/inertia is intriguing...

In information process, a decent analogy is inertia ~ confidence. Inertia is resistance against change, and in an information update, clearly the reason to update a previous state of information which is in contradiction with new information, is directly dependent on the confidence in the contradictory piece, and the confidence in the prior. This ensures stability and overfitting.

This inertia also ensures that variation is small. But it also implies that the variation on microstructures with low inertia will be more violent. At some point the fluctuatios are so large that the microstructure can not be distinguished from random fluctuations.

One can also picture that two communicating inertial systems, will converge in information space, and the idea is also (like Ariel caticha thinks) that spacetime structure, and the distance metrics, can be identified with emergent information measures in some kind of information geometric way.

So spacetime might possibly come as an emergent degrees of freedom as a result of communicating observers, and the measures of spacetime, should then hopefully relate back to probabilistic notions (which I do not take from standard probability theory, I rather think of information theory in terms of combinatorics of distinguishable states, and this leads to a quantized probability itself. Ie. probability doesn't take a continuum of vales from 0 to 1, it´s all about combinatorics, and the continuum probability would only be recovered in the larg number limit, but I think interesting physics happens when the large N approximation is invalid). Ariel CAtichas has hoped to derive GR from a MaxEnt principle where he chooses a particular entropy. See http://arxiv.org/PS_cache/gr-qc/pdf/0508/0508108v2.pdf I don't share his reasoning all the way, but he has many good insights worth reading.

One technical advantage of his, is that it comes with a natural "cutoff" to prevent infinities. The cutoff is simply implicit in the discreteness of the combinatorics. And the complexity of the structures will probably take the role of mass. The mass of the microstructure is then simply the confidence in the microstructure.

But this cutoff is not universal, it would be a function (among other things) of the observers own mass. Thus a massive observer, sees a different discretization spectrum of the "continuum" than a light observer does. As I see it this quest, in the way I picture the solution, really implies a full reconstruction of the mathematical contiuum. The flat introduction of the continuum in physics is a big leap. There is no doubt even a physics behind the continuum. And what you miss out when you flatly introduct contiuum models is that you loose track of the physical degrees of freedom and get lost in the mathematical redundance. Also situations such as infinity - infinity or 0*infinity is highly unphysical, yet they tend to show upp all over the place. This is patological IMHO.

But this is tricky, and I have regular headache due to this. I am trying my best to find what others are doing, and I have some favourites, but it still seems along way to go. But I have at least managed to find hte confidence to stick myself to the quest.

/Fredrik

 

[ConradDJ]

Seriously, information can only be calculated about some other structure. It never exists in a vacuum. But can co-exist with that structure. It is not fundamental in that the underlying structure must exist first before any information can be calculated about it.

I certainly see your point. If we assume that the world consists of some well-defined factual structure, then there seem to be various ways we can pull out and represent ("calculate"?) aspects of that structure as information about the world.

But if well-defined factual structure is something that has to evolve, physically, we may need another way of understanding information. The point I was trying to make above, not very clearly, is that in our ordinary experience we are constantly making sense of the world and gaining information about it, by comparing different kinds of perceptual information in the context of other information stored in memory. We never perceive anything but perceptual information. Of course it's reasonable to believe this information more or less accurately represents the structure of a well-defined reality out there. But an equally valid inference, I think, is that information can be meaningfully defined purely in and through other information.

There is a partial analogy to human language, in that words are defined by other words. But we can also make words meaningful by pointing to things in the physical world: look, that's what we call a "tree."

I think the deep difficulty in fundamental physics might be that physics is a kind of "language" that has no referent outside itself. Maybe no aspect of its structure just "exists" as a well-defined starting-point for everything else -- maybe everything has to make a definable (measurable) difference to something else (that makes a difference to something else, and so on).

So yes, it doesn't make sense to "calculate information from nothing." But the evolutionary idea is that maybe we don't need a well-defined factual reality to start with. If we had a better understanding of how information is actually measured and communicated in the physical world, we might be able to see how that kind of system could have evolved out of mere quantum randomness.

 

[Fra]

So yes, it doesn't make sense to "calculate information from nothing." But the evolutionary idea is that maybe we don't need a well-defined factual reality to start with. If we had a better understanding of how information is actually measured and communicated in the physical world, we might be able to see how that kind of system could have evolved out of mere quantum randomness.

I think one problem is that with "calcuate" we usually think that calculations are deterministic processes, can as a kind of deduction produces a result given input.

PREMISE=STATE1 --/given deterministic rule of reasoning/--> STATE 2

So can we deduce something, as per some deductive rules of reasoning, given NO premises. Of course not.

But, if we instead look at a kind of inductive evolving reasoning, then we can make a guess, a random guess if you like, just to break the symmetry so to speak, and then act upon the feedback. Once things are in motion, it keeps evolving both the states and the reasoning itself.

The obvious problem with induction, and guessing is that it seems to not be scientific at first sight. How do you know if a guess is right or wrong, if there is no clear rules of reasoning, and no clear premises?

I think the answer is to put this in a larger context. Good reasoning, is reasoning that is "successful", bad reasoning is not constructive, and in fact self-destructive. In an evolutinary perspective successful laws, are the one that are consistent with it's own environment.

Look at how people make money on the stock market. How can you make money, without money? how can you make billions, if you have nothing to bet with? Look at how the stock market works, if you have no money, convince your environment that you have good ideas, lend money, take well calculated chances, and if your bettings strategy is among the better, you will make money. Return your loans, and you have created money from nothing.

Here an admittedly silly and oversimplified schematics

0: UNCERTAIN STATE --/UNCERTAIN REASONING/--> ACT
1: FEEDBACK FROM ENVIRONMENT --/UNCERTAIN REASONING/--> REVISE STATE & REASONING
2: GOTO 0

Howto separate the state from the reasoning, is something I have some ideas on but that's not relevant to the main point I think. Somehow, the nature of the uncertainty, SUGGESTS, the reasoning (ie. the way to proceed), pretty much by statistical reasons, you execpt the entropy to increase. But it's more involved and would involved a reconstruction of probability theory and the contiuum.

If you see the analogy, it's not totally unlike how I picture physics. All you need is a fluctuations if you like, and then the ball is rolling, and nothing will stop if from evolving into things that look nothing like the fluctuation it started as.

/Fredrik

 

[friend]

But if well-defined factual structure is something that has to evolve, physically, we may need another way of understanding information.
...
But the evolutionary idea is that maybe we don't need a well-defined factual reality to start with.

Your use of the word evolve seems to imply a direction of time. Time may be an emergent, classical thing; time might not exist as we know it at the quantum level.

I think, is that information can be meaningfully defined purely in and through other information.

I'd have to see an example of what you are talking about.

 

[ConradDJ]

Your use of the word evolve seems to imply a direction of time. Time may be an emergent, classical thing; time might not exist as we know it at the quantum level.

... I'd have to see an example of what you are talking about.

Or, time as we know it might be an evolutionary process. When life first began, it probably wouldn't have been recognizable as life, immediately. When some sort of network of connections first got going in physics, that could evolve in some sense, there might not have been a direction of time definable outside that network.

I've tried to start a thread in the Philosophy section that tries to get at this from another angle -- called "Are Space and Time Definable without Atoms?" I think it's pertinent to the subject of this thread, i.e. can we imagine an evolutionary scenario that's not based on black-hole reproduction.

Besides your very reasonable request for examples, there's a lot in this thread that I want to think about further, as I get some time.

Conrad

 

[ConradDJ]

On "information defined in terms of other information" --

You can think of it in terms of equations -- velocity is defined as distance divided by time. To measure a velocity you need to be able to measure a distance and a time.

Or, you can define a velocity as a momentum divided by a mass, and measure a velocity that way (observe the impact of a body of known mass on another body).

Mathematically, physics is a semantic web of parameters that define each other through equations. Each parameter (like velocity) can be defined in several different ways. Some parameters seem more "primitive" than others -- like distance and time and mass -- because we have rulers and clocks and scales that seem to measure these things "directly". So in classical mechanics it seemed to made sense to consider parameters like velocity, energy and action as in some way secondary, derivative.

On the other hand, Relativity suggests that the velocity of light plays a fundamental role, while space-intervals and time-intervals lose their absolute character. And QM gives us a fundamental constant of action. It remains unclear where mass comes from. So it's no longer obvious which parameters should be considered the fundamental ones, from which the others are derived. What is clear, I think, is that any parameter that could not be defined in terms of the others, in this inter-referential system, would simply have no significance in physics.

Every mathematical definition of a parameter points to a potential way of measuring that parameter. The "semantic" structure of the web of parameters tells us something about the physical structure of the web of interactions -- many different kinds of interactions -- that can actually determine the values of these parameters in a specific case.

We tend to take for granted this inter-referential character of the structure of physics. Of course, every parameter has to be definable in terms of others, to have any meaning. Of course, for every real, physical characteristic of a thing, there has to be some physical context in which it can be observed. A physical characteristic that could not be observed in any way, could not be defined in any meaningful way.

But it's easy to make up examples of systems that don't have any way of defining their own properties. Take a Minimal Newtonian Cosmos, consisting of pointlike mass-particles interacting in Euclidean space and time, only via Newtonian gravity -- nothing else. This seems like a mathematically well-defined system, but clearly it provides nothing that could actually measure a distance or a time-interval or a mass.

So maybe there's something very special about a world like ours, that does provide the physical means for observing and defining every one of its characteristics, in terms of other observable characteristics... that not not only is what is, it but also communicates everything about itself, to itself.

It seems to me that the role of "the observer" in Relativity and especially in QM points us in this direction. QM in particular seems to be telling us that there are no facts about the world except communicated facts. So then, understanding the "semantic structure" of the interaction-web -- how different kinds of facts are defined and communicated in terms of other kinds of facts -- might turn out to be important.

 

[Fra]

Or, time as we know it might be an evolutionary process.

I agree with this. In my use of the word evolve, I make no fundamental distincion between time evolution as we know it, and possible evolution of law. It's just a matter of scales.

The arrow of time is only a sort of preferred direction of change. Why there is a preferred direction of change, and how this preferred direction is defined by each observers is part of the evolving idea.

What we usually think of as say hamiltonian time evolution, can still be thought of as part of a larger evolving process, where the hamiltonian is also evovling it's just that we have imagined a part of the structure as fixed law, to define parameterizing time according to that. I don't there is an objective and uniqe choice in such separation.

I see subjective time as an internal parameterization of the expected changes, consistent with the subjective information. This would define at locally a direction of time.

The relation between different observers, that we usually think of as a symmetry (ie. the idea that reality is indifferent to the choice of observers, and thus reality is represented not by the individual observes view, but with the symmetries relating the set of all observsers) is IMO emergent only. The idea of these symmetries is a typical realist argument IMHO.

My objection is that symmetries are not primary observables. The symmetries are only inferred, by risky arguments, from observations, and then used as a basis for further reasoning - this in itself is sound and good. What is not conceptually sound IMHO at least, is this what I call "realist view" of what a symmetry really is.

It seems to me that the role of "the observer" in Relativity and especially in QM points us in this direction. QM in particular seems to be telling us that there are no facts about the world except communicated facts. So then, understanding the "semantic structure" of the interaction-web -- how different kinds of facts are defined and communicated in terms of other kinds of facts -- might turn out to be important.

I agree with the sentiment here too. This is also rovelli's sentiment, in his Relational QM. He even says that there are only relations, and furthermore only relational relations. However, he does not go all the way. At some point he simply throws in a wet sock and says about "beeing communicated", that communication is a physical interaction that is described by quantum mechanics (ie. the same QM;why?). In this neighbourhood of his reasoning he also notes the wish not to discuss the meaning of probability.

This is exactly (IMHO) where the wet sock is stuck. I would like to revise his reasoning from this poitn and the reconstruct the rest. This is why I couldn't appreciate the logic foundatin of LQG. It really is not radical enough.

/Fredrik

 

[John86]

http://arxiv.org/abs/0906.2700

Anthropomorphic Quantum Darwinism as an explanation for Classicality
Authors: Thomas Durt
(Submitted on 15 Jun 2009)
Abstract: According to the so-called ``Quantum Darwinist'' approach, the emergence of ``classical islands'' from a quantum background is assumed to obey a (selection) principle of maximal information. We illustrate this idea by considering the coupling of two particles that interact through a position-dependent potential. This approach sheds a new light on the emergence of classical logics and of our classical preconceptions about the world. The distinction between internal and external world, the Cartesian prejudice according to which the whole can be reduced to the sum of its parts and the appearance of preferred representation bases such as the position is seen here as the result of a very long evolution and would correspond to the most useful way of extracting stable and useful information from the quantum correlations.

 

[Max™]

Ah, now I get where you were going with this.

The future handshaking with the past gives a present, each observer sees these handshakes from their own perspective, through another series of handshakes, and thus each observer claims a unique set of handshakes to be simultaneous. Their Now is not the same as any other observers Now.

I like to zoom all the way out and consider the state of a Universe as a whole with these processes going on inside of it.

For some reason I get this image of a sphere, which I know to be representing a 4 dimensional object, filled with static. This static is composed of 3-dimensional nows as described by each observer, at each location, at each point in time. Tracing a line between adjacent bits of static can provide a causal relationship divided into spacelike, timelike, and lightlike neighbors.

Each bit of static represents a snapshot of the Universe, and the reason I feel compelled to toss in the word static is that they're jostling each other, constantly blurring and tweaking the snapshots described by their neighbors.

I don't however see a reason why this would have to happen according to the causal rules which we follow, we are composed of certain types of structures, and those happen to "face downstream". If an event we observed and moved beyond were changed, we would not know of it.

The interactions between the now static bits are intimately related to the speed of light.

Any changes induced by the jostling from neighbor bits would be passed along at or below that rate. The act of observation itself would be the same type of effect, and propagate accordingly.

An idea I've been developing, which ties into these concepts, relates the extent to which observers can influence/have their now bits influenced to their mass. I digress on that though, as this is not my thread and not the topic of discussion.

 

[Fra]

http://arxiv.org/abs/0906.2700

Anthropomorphic Quantum Darwinism as an explanation for Classicality
Authors: Thomas Durt
(Submitted on 15 Jun 2009)
Abstract: According to the so-called ``Quantum Darwinist'' approach, the emergence of ``classical islands'' from a quantum background is assumed to obey a (selection) principle of maximal information. We illustrate this idea by considering the coupling of two particles that interact through a position-dependent potential. This approach sheds a new light on the emergence of classical logics and of our classical preconceptions about the world. The distinction between internal and external world, the Cartesian prejudice according to which the whole can be reduced to the sum of its parts and the appearance of preferred representation bases such as the position is seen here as the result of a very long evolution and would correspond to the most useful way of extracting stable and useful information from the quantum correlations.

Thanks for the link John! I'll skim it tomorrow.

While I like some of Zurek's notes, I don't share the overall decoherence view because IMHO it is not a true intrinsic picture, it's rather en external reconstruction of fictive intrinsic pictures. But I'll skim that and see if they have a different view.

/Fredrik

 

[Fra]

Ah, now I get where you were going with this.

Not sure if this was directed to me or some other poster in the thread?

The future handshaking with the past gives a present, each observer sees these handshakes from their own perspective, through another series of handshakes, and thus each observer claims a unique set of handshakes to be simultaneous. Their Now is not the same as any other observers Now.

I like to zoom all the way out and consider the state of a Universe as a whole with these processes going on inside of it.

As I see it, one of the points is that the notion of timless or observer independent "state of the universe" or even state space of the universe is flawed as an scientific abstraction and the use of it is source of problems. This is also partly Smolins view as I read him.

He calls the various timeless laws and state spaces of universes fiction, and I think he is right.

The closest think to the zoomed out picture you mention, is still constrained by a subsystem of the universe - the state space of the entire universe will not fit into this subsystem.

So, the result is that your zoomed out view of the universe as a "whole" it nevertheless subjective. And each observer has their own version. However I think this is aslo exactly why we have interactions between observers.

I think your idea works much better if the observer is very large relative to the system in question. For example particle physics. But even here there are problems, if you try to understand WHY particles have a particular action. This again may force us to emphasise that the actions of a particle might be simple only when seen from the proper inside view.

Your "universe at large" is to me an external view, an in particular a non-physical view that never occurs in nature. Ie. there is no physical system in nature which can encode this view - which is also why it is impossible to understand how such universal action is encoded.

/Fredrik

 

[Fra]

Thanks for the link John! I'll skim it tomorrow.

While I like some of Zurek's notes, I don't share the overall decoherence view because IMHO it is not a true intrinsic picture, it's rather en external reconstruction of fictive intrinsic pictures. But I'll skim that and see if they have a different view.

I skimmed part of the paper and there are traits of their ambition that I do like, however for my taste it's not radical enough. They admit that they don't solve all problems, but since there are other problems that need to be solves as well, and I think when an attempt to do so in the way I picture, the basis for their arguments will be removed, thus I am not so motivated to study some partial ideas, based on premises I don't think will survive anyway.

Indeed I also think in terms of evolutionary processes, BUT IMHO it's inadequate to see this as evolution of a reduced basis in a grand master state space. In fact that is exactly what I object to. I claim that even this "master state space" is evolving, and there exists no inert context in which this can be thought of as simple moving around in a space. Their reasoning contains to much birds view for my taste.

I think one needs to be more radical.

But I agree that the structure, and processes that are seen in nature, must be seen as a result of some kind of evolution with a selection. The question is what the quantitative formalism for this should be.

The complication I take seriously is that this abstraction itself, is subject to the same evolutionary perspective. Ie. I don't think it makes conceptual sense to think that our own science can be distinguished from these same constraints.

As I see it the challange is to find the coherent reasoning which treats physical law on the same footing as the action of a physical system. The microstructure of matter, containing it's actions upon the environment must be a result of the same process as is evolution of physical law, since physical law is really the expectations of the future, given the present, which in turn implies an action, given a sort of "rational player" assumption.

This coherennce in reasoning is missing as I see it in that paper, which is why I think it's not radical enough.

/Fredrik

 

[Max™]

Where is the assumption that the overall state space configuration is unchanging coming from?

I didn't intend that myself, I'd think it would be a strange result to have a state where the components were constantly undergoing changes, but the overall state remained static.

I didn't mean static as in unchanging, I meant like on your TV when you swap to an unused channel. *kksssshhhhhhhhhhhh*

 

[Fra]

Where is the assumption that the overall state space configuration is unchanging coming from?

I admit I found your first post a little confusing in the first place. I made a closest fit assumption to what you meant :)

I interpreted this since you talk about the "state of the universe as a whole". This is a typical phrasing (IMO) for someone picturing fixed state spaces.

Also maybe there is a double confusion, you say

Where is the assumption that the overall state space configuration is unchanging coming from?

I was talking about the configuration state space, not state space configuration. I'm not sure what you mean. I'm not suggesting that you say that universe is static, but I THOUGHT you said that the universe is changing, but you are describing this as a evolution in a picturs statespace, or multiverse if you want (the effect it the same)??

If this is wrong then just forget my comments. I probably got your first post all wrong then :) sorry.

/Fredrik

 

[Max™]

Even with a configuration space, it doesn't have to remain in a single state.

The impression I'd think would be the view that time is unchangeable implies a single configuration state. I was speaking of the opposite, that there should be no single configuration space state allowable in general, as that is almost as bad as implying a preferred frame of reference from the point of view of time.

 

[Fra]

Even with a configuration space, it doesn't have to remain in a single state.

The impression I'd think would be the view that time is unchangeable implies a single configuration state. I was speaking of the opposite, that there should be no single configuration space state allowable in general, as that is almost as bad as implying a preferred frame of reference from the point of view of time.

Oh, then it was me misunderstanding you. That sounds a lot better to me and then we are in closer agreement.

But then I think I understand your post #45 even less. I probably didn't understand your idea.

/Fredrik

 

[Max™]

Just a failure of language to represent ideas which should be explained mathematically in the first place.

 

[Fra]

Just a failure of language to represent ideas which should be explained mathematically in the first place.

Very true. But finding not just some arbitrary mathematics, but the correct mathematics that gives us predictive power is a core part of the problem. My personal view on that is that immature or flawed mathematics can sometimes be far more ambigous and deceptive that plain english becuase it gives the impression of beeing precise, but sometiems arbitrarily so. After all there is a difference between mathematics and physics.

The formalism I'm working on as part of my personal research is something I intend to develop much more before publishing anything, otherwise I KNOW it will come out ambigous. Otherwise you just trade the preciseness problem for the interpretational problem. We need both.

But so far what I'm pondering is a discrete model, where actions and expectations are combinatorically explored. But with the twist that this entire "model, discrete expectaions" etc are not to be thought of as in a realist sense, rather each observer/sub system has their own discrete system; and physical interactions are when several such systems interact. There are several problems still to solve, before I can defend the choice of mathematics. Most important are to show that this scheme solves at least some of the open problems, and correctly can incorporate the standard model. Anything less and the mathematics is not defendable.

/Fredrik